Yes, I believe it's the same issue B1ts reported. My guess would be that there's a bug in the random test generator, because it appears that some random tests expect the wrong answer. However, if that's the case then I don't really get why some people's code is not impacted. Have you tried submitting your code several times (about 10-15)? It may just be that you haven't run enough random tests to find the faulty ones.

Some random cases are wrong; I submitted the same code several times and the 3rd time it passed all the tests, and I can't find any obvious mistakes in it.